State reloading pattern for receiver hooks

I don't know if it's a good idea for the library code to be doing this instead of handling state loading on the user implementation side (like we're doing for fungible tokens). It'll be ok for now since it's going into a dev branch but I think it'll cause problems in actor code that carries additional state of its own beyond the `NFTState`

It does open up one easy optimisation though: we can pass through the original `MintReturn` data in the intermediate struct and simply return that if there's no change once the hook call completes. I've been playing with the idea [here](https://github.com/helix-onchain/filecoin/commit/b0a291b700152d955ae33dc7628fa39925e0a00b) (sending original data) and [here](https://github.com/helix-onchain/filecoin/commit/27a35220b174d75f03aa20f2aff042a502a9f521) (trying to centralise the 'has state changed' check) for the fungible token

_Originally posted by @abright in https://github.com/helix-onchain/filecoin/pull/138#discussion_r1002801864_